131 patient-derived xenograft models were generated for non-small cell lung carcinoma and were profiled at the genome, transcriptome and proteome level by analysis of gene copy number variation, whole exome sequencing, DNA methylation, transcriptome, proteome and phospho(Tyr)-proteome. At the proteome level, the human tumor and murine stroma were discernible. Tumor proteome profiling resolved the known major histological subtypes and revealed 3 proteome subtypes (proteotypes) among adenocarcinoma and 2 in squamous cell carcinoma that were associated with distinct protein-phosphotyrosine signatures and patient survival. Stromal proteomes were similar between histological subtypes, but two adenocarcinoma proteotypes had distinct stromal proteomes. Proteotypes comprise tumor and stromal signatures of targetable biological pathways suggesting that patient stratification by proteome profiling may be an actionable approach to precisely diagnose and treat cancer.